In the Nordic countries, migrants in general and refugees in particular have higher unemployment rates, lower employment rates and are more often in precarious employment compared to natives (Brochmann & Hagelund 2011; OECD 2016). In fact, in Denmark, the employment rate of refugees never reaches the same level as that of the native-born population (Schultz-Nielsen & Skaksen 2017).
In this article, we examine the preferences of Danish employers with regard to the recruitment of jobseekers with a refugee background and test how different country backgrounds affect recruitment preferences. We use a vignette experiment with descriptions of fictitious job applicants to explore whether refugees are disadvantaged on the Danish labour market in the sense that employers are less likely to want to hire people with a refugee background in a recruitment situation.
By manipulating the country of origin of the job applicants, we further explore whether refugees from specific countries (Iraq, Ethiopia and the Ukraine) are more disadvantaged in the recruitment process than refugees with no country of origin specified. Our main contributions to the existing research are the findings that employers are less likely to indicate that they are willing to hire migrants with a refugee background compared to the control situation and that there is an additional ‘country of origin’ penalty (country penalty for short) for Ethiopian jobseekers. The term origin penalty is not a commonly used term in labour market or migration research. In fact, we have not been able to find a single publication using the term. This is also the case for the term national penalty. However, a related and frequently used concept is the term ethnic penalty. The term is used to denote differences in labour market outcomes between natives and members of an ethnic minority population that persist even when controlling for other variables relevant to labour market success (Midtbøen 2015a: 185). In this article, we use the term country penalty because refugees from the aforementioned countries may have different ethnic backgrounds. Setting the article apart from much of the existing research, especially Danish research, we focus on country penalties rather than ethnic penalties and refugee migrant penalties (by which we mean the combined penalty of being a refugee in addition to being a migrant). We find that having a refugee background lowers the employers’ assessment of the candidate. This is especially the case for our vignette with an Ethiopian refugee jobseeker. This candidate has a significantly less positive assessment than a refugee with no country of origin specified (a further country penalty). As illustrated in the literature review, this is in line with current research that finds evidence of ethnic penalties on the labour market. As country penalties are underexplored in Denmark, we focus on refugees from Iraq, Ethiopia and Ukraine (see the review section). We furthermore chose the aforementioned countries because they represent diverse minority groups in Denmark (see the case and design section).
The article is structured as follows. First, we review the existing literature on minority labour market discrimination, primarily focusing on studies using vignettes and field experiments. Second, we present the case, design and methods of the study. Third, we present the findings of our study. We end by concluding and discussing the limitations and possible interpretations of the study as well as possible explanations for the findings.
We review the literature focusing on the role of the employers in relation to recruiting refugees, migrants, ethnic and religious minorities. The explanations for the labour market penalties in this literature tend revolve around the concepts of statistical discrimination and stereotypical discrimination (Midtbøen 2015b; Dahl 2019). Here, an employer negatively evaluates a jobseeker belonging to, for instance, a specific ethnic group or a specific country, because the employer believes that the group as a whole is unproductive, despite not having complete information about the candidate in question.
Non-experimental surveys of employer attitudes generally show positive attitudes towards the recruitment of jobseekers with a refugee background (e.g. Lundborg & Skedinger 2016). Non-experimental attitude surveys are, however, prone to social desirability (Kreuter, Presser & Tourangeau 2008) and thereby fail to capture the actual preferences and intentions of employers (Midtbøen & Rogstad 2012). Field experiments seek to eliminate this type of social desirability bias by measuring the actual behaviour of employers in the form of call-back rates for job interviews. Vignette experiments seek to eliminate the bias by measuring the intentions of the employers by presenting different descriptions of job candidates to random groups of respondents. In this section, we review the literature on employers’ preferences towards recruitment of jobseekers with a refugee background. We limit the literature review to studies on refugees and migrants and primarily include studies using either vignette experiments or field experiments.
Vignette experiments have generally been used to study the recruitment preferences of employers, for instance, with regard to the long-term unemployed (e.g. Van Belle et al. 2018), people with disabilities (Nota et al. 2014), youth unemployment (Hyggen 2017), gender discrimination (Kübler, Schmid & Stüber 2018), ethnic minorities and refugees. Auer et al. (2019) used a vignette experiment among Swiss employers to explore ethnic discrimination in high-skilled and low-skilled jobs. The authors found that jobseekers with an immigrant background are more disadvantaged in high-skilled jobs than in low-skilled jobs. In the same context, Fossati et al. (2017) explored the hiring preferences of Swiss human resource managers using a factorial survey experiment. They found that second-generation job applicants with Turkish or Polish parents are evaluated less positively than Swiss and Spanish job applicants.
Field experiments, also known as correspondence experiments, measure the actual behaviour of employers by sending out fictitious job applications to real job openings. The name of the fictitious job applicant (or other characteristics) was randomised and call-back rates for job interviews were subsequently compared. The majority of these studies found evidence of ethnic discrimination (Blommaert, Coenders & van Tubergen 2014; Carlsson & Rooth 2007; Midtbøen 2016). Zschirnt and Ruedin (2016) conducted a statistical meta-analysis of (almost) all correspondence tests (field experiments) from 1990 to 2005, examining the discrimination rates across OECD countries and across studies. They found that minority job applicants must send 50% more applications than majority groups to receive a call back for a job interview. In a meta-analysis of 97 field experiments, Quillian et al. (2019) explored whether discrimination is more prevalent in some countries than others and whether some groups are more discriminated than others. Most importantly, they found that white immigrant groups are less discriminated than other immigrant groups. In a similar vein, a cross-national field experiment (the GEMM study) examined differences in discrimination rates between Germany, Norway, the Netherlands, Spain, the United Kingdom and the United States (Lancee 2019). The author found that ethnic minorities have a lower chance to be invited to an interview than the majority population, and Germany has the lowest discrimination levels. A Canadian field experiment examined whether there are differences between the call-back rates of small and large firms (Banerjee, Reitz & Oreopoulos 2018). The authors found that call-back rates are lower for Asian applicants in small, medium and large firms, but the differences are not as pronounced in large firms, indicating that large firms may discriminate to a lesser degree.
In a double-comparative cross-national harmonised field experiment, Di Stasio et al. (2019) investigated whether there is a penalty for migrants originating from countries with Muslim majorities and whether there is an additional religious penalty for signalling Muslim faith in a job application. The authors found evidence of both a distinct ethnic penalty and a religious penalty in three out of five countries. To some extent, jobseekers originating from countries with Muslim majority populations are regarded as ‘Muslim by default’ entailing a disadvantage in the job search.
We managed to find a few Danish field experiments of employer discrimination against ethnic minorities. An older field experiment by Hjarnø and Jensen (1997) found evidence that young people with a migrant background are less likely to be invited to a job interview than young native Danes. Villadsen and Wulff (2018) found evidence that both public- and private-sector employers discriminate in the recruitment process. Guul, Villadsen and Wulff (2019) investigated discrimination in the recruitment process among Danish public schools through a field experiment. They explored whether high-performing schools (in terms of a high grade point average) discriminate less than poorly performing schools. They found that discrimination in public organisations is more likely to take place in poorly performing organisations. Another recent Danish field experiment by Dahl and Krog (2018) explored gender differences in call-back rates between job applicants with a Danish sounding name and a Middle Eastern sounding name. Dahl and Krog found larger differences among males than females but overall, job applicants with a Middle Eastern sounding name have a lower call-back rate than those with a Danish sounding name. Dahl (2019) focused specifically on women. In the experiment, identical CVs and job search applications were sent for real job openings. The control and treatment in the experiment were differences in the CV photos and names of the applicants. Dahl found that female minority women must send 18% more job applications to receive a callback for a job interview compared to Danish women. In addition, Dahl found that minority women wearing a headscarf, signalling Muslim faith, must send 60% more applications than the control group to receive a callback.
The available vignette studies and field experiments indicate that ethnic minority groups, migrants and descendants from certain countries, and job applicants signalling Muslim faith are disadvantaged in the recruitment process. The aforementioned groups are less likely to be invited for job interviews than the majority population and the penalty differs by origin country, job type and company size. We only found a few studies of Denmark, and no studies examining employers’ preferences towards hiring different refugee groups. Our study thus contributes to the existing literature by exploring whether employers are less likely to indicate willingness to recruit refugees and whether their assessments of refugees from different countries differ.
Danish employers and workplaces are interesting to study in relation to the employment of refugees. First, while the employment rate in Denmark is among the highest in the world at 74% compared to an OECD average of 64.2% (OECD 2021), there is a large gap of roughly 20 percentage points between the employment rate of immigrants and refugees from non-EU countries and native Danes (Statistics Denmark [ras200]1). Second, the regulation of the dismissal and recruitment decisions of Danish companies is limited and flexible, which means that employers are free to decide whom to recruit and dismiss (Bredgaard, Larsen & Madsen 2006). Third, the wage structure is compressed with high minimum wages, and skills demands are high, which disfavours low-skilled and un-skilled jobseekers (Brochmann & Hagelund 2011). Denmark may be considered a case where the labour market integration of refugees is difficult. This is supported by Danish data showing that only 37% of refugees are in employment 3 years after they were granted a residence permit.2
In this study, we combine an experimental design and survey methodology, a vignette experiment (Aguinis & Bradley 2014). Vignette experiments measure hypothetical situations and intended behaviour (Aguinis & Bradley 2014). This is a downside compared to field experiments that measure the actual behaviour of employers by examining call-back rates for fictitious job applicants for real job openings. There are, however, several arguments for conducting survey vignette experiments. First, it is more cost efficient to distribute an online survey with a vignette experiment included than to apply for several hundred real jobs. Second, there are many ethical considerations. Respondents consent to participate in a survey. Field experiments lack informed consent (Desposato 2018). In addition, employers use time and resources to assess job applicants; in a field experiment, they are, in a way, deceived by being sent false job applications. Therefore, survey vignette experiments are easier ethically to defend than field experiments. The internal validity of our study is strong because of the random assignment into control and treatment groups. Nonetheless, vignettes have lower external validity than field experiments, because they measure intentions to hire (preferences) rather than actual behaviour.
In June 2019, we conducted a survey among a representative sample of Danish workplaces with at least three employees. The survey included the experiment vignette as one of the first questions (see Table 1 for a full description of the vignette) as well as questions concerning corporate social responsibility, attitudes towards national integration policies, recruitment patterns and collaboration with the public employment service. The sample of workplaces was drawn from the CVR-register in which all Danish businesses are listed. A net sample of 5,000 private workplaces and public workplaces was drawn from the register including information about the number of employees, whether the workplace was in the public or private sector and which industry it operated in across ten categories. Workplaces with less than three employees were excluded from the sample. Invitations to participate in the survey were sent to the employers or workplace managers responsible for recruitment by electronic post. Subsequently, the workplaces were called by phone and given the option of answering the survey through a link sent by email or by being guided through the questions by an interviewer. A maximum of two follow-up emails (reminders) were sent to each potential respondent. The response rate was 42.9% (1,977 valid responses amounting to answers from 1,977 different workplaces), which is relatively high for Danish employer surveys. Among the valid responses, 55% of the respondents were the employers themselves and the remaining 45% were workplace managers or recruitment personnel. The survey is representative of the concerning industry, but larger workplaces are slightly overrepresented among those who answered the survey (the mean number of employees is 27.4 among those who answered the survey, compared to 23.3 among the sampled workplaces).
|Control group||Imagine a situation where you are about to hire a new employee. A person has applied for the job and tells you that she/he has relevant work experience. The applicant is positive, energetic and a hard worker.|
|Treatment 1||Same as control + The person tells you that she/he fled her/his country of origin 2 years ago.|
|Treatment 2||Same as control + The person tells you that she/he fled Iraq 2 years ago.|
|Treatment 3||Same as control + The person tells you that she/he fled Ethiopia 2 years ago.|
|Treatment 4||Same as control + The person tells you that she/he fled Ukraine 2 years ago.|
As mentioned, the survey vignette experiment was conducted in the first part of the survey to measure the recruitment preferences of employers with regard to jobseekers with a refugee background. A survey experiment mimics a randomised controlled experiment by assigning subjects to control and treatment groups (Angrist & Pischke 2015). The respondents were randomly assigned to one of five groups. They were presented with a vignette describing a potential job applicant and were asked to rate the likelihood that they would hire the candidate for a vacant position at the workplace. Each group was presented with only one vignette and had no knowledge of the other vignettes. The choice to present respondents with only one vignette was made because including too many vignettes can lead to fatigue and information overload in respondents (Aguinis & Bradley 2014). Furthermore, the total number of questions in the survey had to be limited to prevent fatigue and dropouts from the survey. A practical decision was made to allow for other questions. The downside to this methodological choice is that multiple vignettes would have allowed us to investigate several causal factors influencing the workplace assessment of candidates. In addition, presenting employers with only one vignette involves a risk of a not excluding a confounding variable (unobserved employer characteristics) or unobserved variable bias. For instance, some employers might have rated all vignettes more positive than others. However, if we, for instance, had presented all employers with the control vignette or an additional treatment vignette, the respondents might become aware of the experiment, introducing a risk of social desirability bias. For these reasons, we chose a single vignette per respondent.
Our article is mainly an empirical paper. We do not set out to test a specific theory in the article but used theory and empirical research in the design of the vignettes. The theory of social distance by Akerlof (1997) (not to be confused with social distancing to prevent the spread of disease) was used as inspiration in the design of the vignettes. The theory states that choices are not made by utility maximising individuals in a social vacuum, but are instead affected by the social context and conformity to norms. This implies that individuals prefer the company of others who are socially close to themselves and avoid interactions with individuals whom they are socially distant from. In terms of employer preferences, this implies that employers favour applicants whom they believe they are socially close to and disfavour applicants they are socially distant from. Religious, racial and cultural differences, and unfamiliarity with a group are thought to increase perceived social distance. In the vignettes, we manipulated the country of origin with the aim of varying perceived social distance from low to high (see the control and treatment groups section). In addition, we tried to incorporate some factors in the vignettes to make candidates attractive to employers. The existing research on popular perceptions of refugees showed that refugees are often more positively perceived than migrants in general (Wyszynski 2018), as they are not responsible for their situation of having to flee their native country, according to deservingness theory (Bjånesøy 2019; Van Oorschot 2006). Signalling a refugee background might increase perceived deservingness of a job, thereby increasing the employer’s assessment of the candidate.
The existing empirical research (eg. Bredgaard & Thomsen 2018; Auer 2018) found that poor language skills, a lack of relevant work experience and even work motivation are three supply-side barriers to the labour market integration of refugees. Building on this research, we tried to signal basic Danish language skills, relevant work experience and work motivation in the design of our vignettes.
The respondents were asked to rate the likelihood that they would hire the candidate from the vignette on an ordinal scale of 1–5 from ‘very likely’ to ‘very unlikely’. The vignette texts are listed in Table 1.
The design allows us to (1) estimate the effect of having a refugee migrant background on the employer assessment of candidates and (2) estimate the effect for refugees from different countries of origin. It is important that respondents consider the job applicants as realistic. Our considerations about the vignette descriptions of the control and treatment groups are therefore outlined in more detail below.
The vignette for the respondents in the control group is a job applicant with generally but not overly positive traits (see Table 1). The job applicant is characterised as having relevant work experience, being positive, energetic and a hard worker. Any information about country of origin and additional background information is left out in the control vignette. We presume that respondents think of a native Dane. We considered supplying the respondents with information about relevant educational attainment but decided to discard this information to ensure that the vignettes for the treatment groups would be considered realistic by employers and recruitment personnel across industries and job functions, because unskilled work does not require a formal education.
In treatment group 1, we present the respondents with the same information about the job applicant as the control group but add the information that the job applicant fled his/her country of origin two years ago (the ‘generic refugee’). The cut-off point of two years was chosen to ensure that the person in the vignette would be assessed to be a rather recent refugee but also for a long enough time to have passed for the job applicant to have acquired some basic Danish skills. As stated in the theory of social distance, we expect that having a refugee background could be a disqualifying factor in the recruitment process, but it could also be a compensating factor according to deservingness theory. In the remaining treatment groups, we test for the implications of the country of origin of the refugees.
In treatment group 2, we add information that the job applicant fled Iraq two years ago. With inspiration from the theory of social distance, we chose Iraq for several reasons. After Turks and Syrians, Iraqis are the third largest minority group with 33,000 Iraqis living in Denmark. (Statistics Denmark [FOLK1C]). Historically, many refugees from Iraq came to Denmark during the early and mid-1990s and again in the 2000s after the second Gulf War. We thus expect respondents to be relatively familiar with Iraqis in Denmark. Theoretically, we expect that familiarity will decrease perceived social distance. Implicitly, this treatment vignette also signals a person with a Muslim and Middle-Eastern background, which several studies have found to be a disqualifying factor (see the review section). This is thought to increase social perceived social distance. Overall, we designed this vignette with the intention of creating a group representing a mid-level social distance to employers.
In treatment group 3, we add information about the job applicant having fled Ethiopia two years ago. Ethiopians are a small minority group with 1,900 Ethiopians living in Denmark. Because of the Ethiopian civil war (1974–1991), Denmark received 849 Ethiopian refugees from 1983–1992 of which many were granted family reunification (Statistics Denmark 2008). Denmark also experienced a surge in refugees from Ethiopia from 2014 to 2017, granting asylum to more than 100 Ethiopians during this period. This is a rather large number compared to the number of Ethiopians living in Denmark. Furthermore, Ethiopia is one of the poorest countries in the world (World Bank Group 2015). We therefore expect respondents to associate Ethiopians with people with a dark skin colour coming from an economically under-developed country. In addition, we presume that respondents have limited knowledge about the religious background of Ethiopians and of Ethiopians in general. The Shades of brown research examined discrimination based on skin-tone among African-Americans (Jones 2000). In essence, the findings are that the darker the skin-tone, the greater disadvantagement people face. We incorporated these factors with the intention of maximising perceived social distance, and we therefore expected that the Ethiopian candidate will get rather low ratings from the employers.
Finally, treatment group 4 is presented with a job applicant who fled Ukraine two years ago. Ukraine is an example of an Eastern European country in conflict with Russia and we expect employers to be aware of this. There are 14,000 Ukrainians living in Denmark, most of which are labour migrants according to Statistics Denmark. Each year more than 1,000 Ukrainians are given a residence permit to work in Denmark (Statistics Denmark [VAN5RKA]). From 2014 to 2019 an average of 42 Ukrainians applied for asylum in Denmark annually (Statistics Denmark [VAN66KA]. Most Ukrainians in Denmark are thus labour migrants, but a small minority has a refugee background. We therefore believe that employers will regard the vignette as realistic. The fact that Ukrainians are a well-established labour migrant group in Denmark might signal to employers that Ukrainians, as a group, are motivated and hard working. We expect that most respondents associate Ukrainians with a Christian background and a light skin colour. The Ukrainians thus represent a group with a similar appearance to most Danes a reputation for being hard working. This vignette was designed to minimise perceived social distance, and we therefore expected a rather positive rating from employers.
The differences between the control group and the treatment groups on background variables are rather small, suggesting that the randomisation process was successful in creating similar groups (see Table 5 in the Appendix). The results of the vignette experiment are shown in Table 2. The total number of respondents (workplaces) is 1,977 with roughly 400 respondents in each of the five groups.
|Very unlikely to hire||Unlikely to hire||Neither likely or unlikely to hire||Likely to hire||Very likely to hire||Total (n)||Mean (scale of 1–5)|
|Treatment 1 (generic refugee)||11.2%||8.3%||29.9%||28.7%||21.9%||100% (421)||3.42|
|Treatment 2 (refugee from Iraq)||16.4%||10.3%||26.2%||24.4%||22.7%||100% (397)||3.27|
|Treatment 3 (refugee from Ethiopia)||17.9%||9.5%||34,0%||21.6%||17.1%||100% (380)||3.11|
|Treatment 4 (refugee from Ukraine)||13.9%||11.0%||26.7%||30.8%||17.7%||100% (390)||3.27|
In relation to Table 2, it is worth noting that, across groups, only a small minority of workplaces indicate that they are either ‘unlikely’ or ‘very unlikely’ to want to hire the candidate from the vignette. This share is lowest in the control group (roughly 11%) and for the treatment groups it is higher, ranging between 18% and 27%. Therefore, very few workplaces have a negative assessment of the candidates, regardless of refugee background. Across groups, there is also a rather large share of respondents giving neither positive nor negative assessments (neither likely nor unlikely to hire the candidate). The share is lowest in the control group (20%) and highest in treatment group 3 (refugee from Ethiopia). Turning to the share of workplaces that indicate that they are ‘likely’ or ‘very likely’ to hire the candidate we see much more pronounced differences. The overwhelming majority, approximately 70% of the workplaces in the control group, state that they are likely or very likely to hire the candidate. The share drops considerably for all treatment groups. For treatment groups 1, 2 and 4 (generic refugee, refugee from Iraq and refugee from Ukraine) it drops by roughly 20 percentage points. For treatment group 3 (refugee from Ethiopia), the share drops even further by 30 percentage points. This suggests that treatment group 3 (Ethiopian refugee) is the least likely to be hired among the presented candidates. Our results indicate that there is a ‘refugee migrant penalty’ in the sense that workplaces are less inclined to indicate willingness to hire all treatment groups This is supported by the last column of Table 2, showing the means on a scale from 1 to 5. The score (likelihood) is highest for the control group, lower for all treatment groups and lowest for treatment group 3 (Ethiopian refugee). The mean for treatment group 3 (Ethiopian refugee) is lowered primarily by a larger share of employers indicating neutral assessments (neither likely nor unlikely to hire) and by a larger share signalling that they are unlikely to hire the candidate. Overall, it would seem that, when a candidate is a migrant with a refugee background, the inclination to hire the candidate is lowered on average.
To explore whether the differences between groups are statistically significant, we performed a pairwise comparison of means using the Šidák multiple comparison method (Table 3).
|Contrast||p-value (Sidak)||95% confidence interval (Sidak)|
|Generic refugee vs. control||–0.461||0.000||–0.710||–0.212|
|Refugee from Iraq vs. control||–0.612||0.000||–0.864||–0.360|
|Refugee from Ethiopia vs. control||–0.774||0.000||–1.029||–0.519|
|Refugee from Ukraine vs. control||–0.605||0.000||–0.858||–0.351|
|Refugee from Iraq vs. generic refugee||–0.151||0.599||–0.399||0.097|
|Refugee from Ethiopia vs. generic refugee||–0.313||0.005||–0.563||–0.062|
|Refugee from Ukraine vs. generic refugee||–0.144||0.672||–0.392||0.105|
|Refugee from Ethiopia vs. refugee from Iraq||–0.162||0.538||–0.416||0.092|
|Refugee from Ukraine vs. refugee from Iraq||0.007||1.000||–0.245||0.260|
|Refugee from Ukraine vs. refugee from Ethiopia||0.169||0.480||–0.086||0.424|
In Table 3, we investigate the between group differences, and explore how large the negative effect is of country of origin and having a refugee background. Compared to the control group (no country of origin specified) being a refugee lowers the score around half a point (0.461) on the scale of 1–5. The effect is significant at a 0.000 level. Compared to the control group, the score is lowered by 0.612 points for treatment group 2 (refugee from Iraq), by 0.774 points for treatment group 3 (refugee from Ethiopia) and by 0.605 points for treatment group 4 (refugee from Ukraine). In practice, this means that refugees from all three countries get worse evaluations by the employers than our control group consisting of native Danes. All three results are significant at a 0.000 level, which implies that the results in this sample can be generalised to the population of Danish employers. Contrary to our initial assumption, having a Ukrainian background did not alleviate the negative effect from having a refugee background in our vignettes. We believe that the absence of differences between refugees from Ukraine and Iraq is an important non-finding, in the sense that this was an unexpected finding contrary to our initial assumption. As stated earlier, we developed the vignette with the Ukrainian refugee jobseeker with the following intentions. We wanted to give the respondents a mental cue to think of a person from a mainly Christian country, with an appearance, customs and work ethics similar to most Danes. How the vignette was interpreted by the respondents is a matter for discussion. The vignette with the refugee jobseeker from Iraq was developed to provide respondents with a cue to think of a person from a mainly Muslim country but also to represent a group many Danes are familiar with. Based on Di Stasio et al. (2019) and Dahl (2019) we expected the vignette with the jobseeker from Iraq to be evaluated less positively than the vignette with the Ukrainian refugee jobseeker. However, this was not the case.
It is worth noting that the vignette with the Ethiopian refugee jobseeker (treatment group 3) was the only vignette evaluated significantly less positively than the ‘generic refugee’ with no country of origin specified. In the vignettes, both refugee jobseekers from Iraq and from Ukraine are given less positive assessments than the control group but not less positive than the ‘generic refugee’. We hypothesised, that the Ethiopian refugee would be perceived as the socially distant to natives Danes.
The results might be driven by an employer preconception that refugees have poor language skills in general, regardless of their country of origin. In relation to this, it is interesting if effects differ for workplaces indicating a low level of language skills requirements. We asked all respondents (workplaces) about the necessity of the Danish skills to work at the workplace, using the following question ‘How necessary is it to be able to understand Danish to work at the workplace?’ Response categories were ‘very necessary’, ‘not very necessary’, ‘not necessary at all’ and ‘don’t know’. To test for heterogeneous effects we performed the analyses on a subset of the respondents, including only those respondents who indicated that it is ‘not very necessary’ or ‘not necessary at all’ to be able to understand Danish to work at the workplace (Table 4).
|Very unlikely to hire||Unlikely to hire||Neither likely or unlikely to hire||Likely to hire||Very likely to hire||Total (n)||Mean|
|Treatment 1 (generic refugee)||5.0%||6.3%||16.3%||38.8%||33.8%||100%
|Treatment 2 (refugee from Iraq)||3.8%||8.9%||20.3%||29.1%||38.0%||100%
|Treatment 3 (refugee from Ethiopia)||6.6%||6.6%||31.9%||30.8%||24.2%||100%
|Treatment 4 (refugee from Ukraine)||4.8%||2.9%||22.9%||36.2%||33.3%||100%
The subset of employers indicating limited requirements for Danish language skills include 433 respondents (workplaces), implying that statistical power weakens. Therefore, we might not be able to identify if one of the treatment groups differs from the control group. However, from the estimated means in Table 4, there should not be cause for concern. The means of treatment groups 1, 2 and 4 are all around 3.9 – very close to the mean of our control group (4.08). In the vignette, it thus appears that the ‘generic refugee’ and the refugees from Iraq and Ukraine are not disqualified in the recruitment processes of workplaces where the ability to understand Danish is not necessary for working at the workplace. The results indicate that having a refugee background is not a disqualifying factor in workplaces with low language skill requirements. One exception, however, is the refugee from Ethiopia (treatment group 3). The mean in Table 4 is roughly 0.5 lower on the scale of 1–5 than the mean of the control group. To test whether there is a significant difference in the mean between the control group and treatment group 3, we performed a pairwise comparison of means using the Šidák multiple comparison method as in Table 3 (see Table 6 in the appendix). We find that the negative effect for the Ethiopian refugees in the vignette persists even when focusing only on workplaces with limited language skills requirements. This indicates that Ethiopian refugees are disadvantaged in the job search process compared to native Danes, even in workplaces with few language skills requirements. On the positive side, we see that the majority of workplaces with low language skills requirements (55%) report that they are ‘likely’ or ‘very likely’ to hire the Ethiopian jobseeker in the vignette.
To further investigate heterogeneous effects, we performed the same analyses including the following subgroups of respondents (1) in the public and private sector (see Table 7 in the appendix), (2) with less and more than 10 employees (see Table 8 in the appendix) and (3) workplaces that are and are not members of an employer association (see Table 9 in the appendix). These subgroup analyses yielded qualitatively the same results as Table 3, indicating that having a refugee background is a disqualifying factor in the recruitment process.
As described previously, we found evidence of a refugee migrant penalty through our vignette experiment. Building on research, which finds that vignette experiments tend to under estimate labour market penalties, we expect that the negative effects of having a refugee migrant background will be amplified in a real-life situation (Wulff & Villadsen 2019).
This relates to the validity of our study and vignette experiments more generally. As mentioned earlier, the internal validity of our study is strong because of random assignment to control and treatment groups. However, our study has some limitations worth mentioning here. A general critique of vignette experiments is that it is much less costly for employers to indicate willingness to hire a candidate in a survey than to actually hire a person in a real-life situation (external validity). Furthermore, because of the design, we are not able to empirically explain why the treatment vignettes are evaluated less positively than the control vignettes. For instance, is the vignette with the Iraqi refugee jobseeker evaluated less positively than the control vignette, because Iraq is a mainly Muslim country? Is the vignette with the Ethiopian refugee jobseeker evaluated less positively than both the control vignette and the first-treatment vignette (the ‘generic refugee’), because employers are not familiar with people with an Ethiopian background, because they think of a person with dark skin colour or because they believe that Ethiopia in an economically under-developed country? Further research is needed to provide these explanations.
An additional limitation of our study is that we cannot estimate the isolated effect of having a refugee background. This would have required us to compare two vignettes with two jobseekers from the same country, one with a refugee background and one without a refugee background.
In relation to the design of the vignettes, it is worth mentioning that we did not specify from where the job applicants in the vignettes got their relevant work experience. The employers might have interpreted it as relevant work experience from their country of origin. If this is the case, it would imply that employers devalue work experience obtained in foreign countries.
We examined employer preferences with regard to the recruitment of refugees on the Danish labour market using a vignette experiment that includes five vignettes describing candidates applying for a vacant position. In the survey, employers assessed the likelihood that they would hire the candidate in question. The vignette control group consisted of a rather positively described candidate for the job, with no mention of ethnicity or refugee background, signalling a native Dane. In the first treatment group, we added information about the candidate having a refugee background, in addition to the same rather positive description of the candidate, but no country of origin was mentioned (a ‘generic refugee’). In the three remaining treatment groups, we varied the country origin (Iraq, Ethiopia and Ukraine) in addition to having a refugee background. Overall, we find that the vignettes with refugee job applicants are rated less positively than the control vignette.
We believe that our findings have produced evidence that migrants with a refugee background face labour market penalties in the job search process. Our claim is thus that our findings can be generalised to conclude that refugees face a real disadvantage in the real job search process and not just in the hypothetical situation we examined through the vignettes. This is the case for all three countries of origin (Iraq, Ethiopia and Ukraine). Furthermore, we find that the workplaces evaluate the vignette with the Ethiopian refugee jobseeker significantly less positive than the refugee without country of origin specified (a ‘generic refugee’).
In workplaces with no or low Danish language proficiency requirements, the negative effect on the employers’ assessments of candidates disappeared for all treatment groups except for the refugee job applicant from Ethiopia. Further research is needed to examine why having an Ethiopian background is a disqualifying factor – even in these workplaces.
Our main and contribution to the literature is the identification of a ‘refugee migrant penalty’. Applicants with a refugee background are significantly less positively assessed than the control group. We believe that this finding can be generalised to real-world situations implying that refugees face a penalty on the Danish labour market.
Our second contribution relates to our non-finding of further country of origin penalties for refugee jobseekers from Iraq and Ukraine. In the treatment groups presented with additional information about the country of origin of the refugee job applicant, we find no significant differences between being a generic refugee (with no country of origin specified) and being a refugee from either Iraq or Ukraine. Coming from either of these two countries does not seem to be a further disqualifying (or compensating) factor when applying for a job in a Danish workplace. In relation to the Iraqi jobseeker with a refugee background, this is a rather surprising result because the existing research finds a negative effect of having a Middle-Eastern background or signalling Muslim faith on the likelihood of employment (Di Stasio et al. 2019; Dahl 2019). The vignette with the Ukrainian jobseeker was chosen to signal a person similar to native Danes in terms of appearance and religious background. However, this did not alleviate the negative effects of having a migrant refugee background. This is somewhat contrary to the existing research that finds that discrimination against white immigrant groups is limited (Lancee, 2019). A plausible interpretation of this is that in a recruitment situation, having a refugee background is more of a decisive factor than being similar in several ways to the majority population.
The authors have no competing interests to declare.
Aguinis, H and Bradley, KJ. 2014. Best practice recommendations for designing and implementing experimental vignette methodology studies. Organizational Research Methods, 17(4): 351–371. DOI: https://doi.org/10.1177/1094428114547952
Auer, D. 2018. ‘Language roulette–The effect of random placement on Refugees’ labour market integration. Journal of Ethnic and Migration Studies, 44(3): 341–362. DOI: https://doi.org/10.1080/1369183X.2017.1304208
Auer, D, Bonoli, G, Fossati, F and Liechti, F. 2019. The matching hierarchies model: Evidence from a survey experiment on employers’ hiring intent regarding immigrant applicants. International Migration Review, 53(1): 90–121. DOI: https://doi.org/10.1177/0197918318764872
Banerjee, R, Reitz, JG and Oreopoulos, P. 2018. Do large employers treat racial minorities more fairly? An analysis of a Canadian field experiment data. Canadian Public Policy, 44(2): 1–12. DOI: https://doi.org/10.3138/cpp.2017-033
Bjånesøy, LL. 2019. Effects of the refugee crisis on perceptions of Asylum seekers in recipient populations. Journal of Refugee Studies, 32(1): i218–i237. DOI: https://doi.org/10.1093/jrs/fey070
Blommaert, L, Coenders, M and van Tubergen, F. 2014. Discrimination of Arabic-named applicants in the Netherlands: An internet-based field experiment examining different phases in online recruitment procedures. Social Forces, 92(3): 957–982. DOI: https://doi.org/10.1093/sf/sot124
Bredgaard, T, Larsen, F and Madsen, PK. 2006. Opportunities and challenges for flexicurity—The Danish example. Transfer: European Review of Labour and Research, 12(1): 61–82. DOI: https://doi.org/10.1177/102425890601200107
Bredgaard, T and Thomsen, TL. November 2018. Integration of refugees on the Danish labor market. Nordic Journal of Working Life Studies, 8: 7–26. DOI: https://doi.org/10.18291/njwls.v8iS4.111161
Brochmann, G and Hagelund, A. 2011. Migrants in the Scandinavian welfare states—The Emergence of a Social Policy Problem. Nordic Journal of Migration Research, 1(1): 13–24. DOI: https://doi.org/10.2478/v10202-011-0003-3
Carlsson, M and Rooth, DO. 2007. Evidence of ethnic discrimination in the Swedish labor market using experimental data. Labour Economics, 4(spec. iss): 716–729. DOI: https://doi.org/10.1016/j.labeco.2007.05.001
Dahl, M. 2019. Detecting discrimination. How group-based biases shape economic and political interactions: Five empirical contributions. PhD Dissertation, Department of Political Science, University of Copenhagen.
Dahl, M and Krog, N. 2018. Experimental evidence of discrimination in the labour market: intersections between ethnicity, gender, and socio-economic status. European Sociological Review, 34(4): 402–417. DOI: https://doi.org/10.1093/esr/jcy020
Desposato, S. 2018. Subjects and scholars’ views on the ethics of political science field experiments. Perspectives on Politics, 16(3): 739–750. DOI: https://doi.org/10.1017/S1537592717004297
Di Stasio, V, Lancee, B, Veit, S and Yemane, R. 2019. Muslim by default or religious discrimination? Results from a cross-national field experiment on hiring discrimination. Journal of Ethnic and Migration Studies, 47: 1305–1326. DOI: https://doi.org/10.1080/1369183X.2019.1622826
Guul, TS, Villadsen, AR and Wulff, JN. 2019. Does good performance reduce bad behavior? Antecedents of ethnic employment discrimination in public organizations. Pulic Administration Review, 79(5): 666–674. DOI: https://doi.org/10.1111/puar.13094
Hyggen, C. 2017. Etterlater arbeidstrening arr hos unge ledige? Et vignett-eksperiment av arbeidsgiveres beslutninger ved ansettelser av unge i Norge. Søkelys på arbeidslivet, 34(04): 236–251. DOI: https://doi.org/10.18261/issn.1504-7989-2017-04-01
Jones, T. 2000. Shades of brown: The law of skin color. Duke Law Journal, 49(6): 1487–1558. DOI: https://doi.org/10.2307/1373052
Kreuter, F, Presser, S and Tourangeau, R. 2008. Social desirablity bias in CATI, IVR, and web surveys-the effects of mode and question sensitivity. Public Opinion Quarterly, 72(5): 847–865. DOI: https://doi.org/10.1093/poq/nfn063
Kübler, D, Schmid, J and Stüber, R. August 2018. Gender discrimination in hiring across occupations: A nationally-representative vignette study. Labour Economics, 55: 215–229. DOI: https://doi.org/10.1016/j.labeco.2018.10.002
Lancee, B. 2019. Ethnic discrimination in hiring: Comparing groups across contexts. Results from a cross-national field experiment. Journal of Ethnic and Migration Studies. DOI: https://doi.org/10.1080/1369183X.2019.1622744
Lundborg, P and Skedinger, P. 2016. Employer attitudes towards refugee immigrants: Findings from a Swedish survey. International Labour Review, 155(2): 315–337. DOI: https://doi.org/10.1111/ilr.12026
Midtbøen, AH. 2015a. Ethnic penalties in Western labour markets-contributions, explanations, critiques. Nordic Journal of Migration Research, 5(4): 185–193. DOI: https://doi.org/10.1515/njmr-2015-0022
Midtbøen, AH. 2015b. Etnisk diskriminering i arbeidsmarkedet. Tidsskrift for Samfunnsforskning, 55(1): 3–30. DOI: https://doi.org/10.18261/ISSN1504-291X-2015-01-01
Midtbøen, AH. 2016. Discrimination of the second generation: Evidence from a field experiment in Norway. Journal of International Migration and Integration, 17(1): 253–272. DOI: https://doi.org/10.1007/s12134-014-0406-9
Midtbøen, AH and Rogstad, J. 2012. Discrimination-methodological controversies and sociological perspectives on future research. Nordic Journal of Migration Research, 12(3): 203–212. DOI: https://doi.org/10.2478/v10202-011-0046-5
Nota, L, Santilli, S, Ginevra, MC and Soresi, S. 2014. Employer attitudes towards the work inclusion of people with disability. Journal of Applied Research in Intellectual Disabilities, 27(6): 511–520. DOI: https://doi.org/10.1111/jar.12081
OECD. 2021. Employment Rate OECD Data. Available at https://data.oecd.org/emp/employment-rate.htm.
Quillian, L, Heath, A, Pager, D, Midtbøen, AH, Fleischmann, F and Hexel, O. 2019. Do some countries discriminate more than others? Evidence from 97 field experiments of racial discrimination in hiring. Sociological Science, 6: 467–496. DOI: https://doi.org/10.15195/v6.a18
Schultz-Nielsen, ML and Skaksen, JR. 2017. Indvandreres uddannelse. ROCKWOOL Fondens Forskningsenhed, Arbejdspapir nr. 48 [Educational attainment of immigrants]. Available at www.rockwoolfonden.dk.
Statistics Denmark. 2008. Indvandrere i Danmark. [Immigrants in Denmark], Statistics Denmark. Available at https://www.dst.dk/Site/Dst/Udgivelser/GetPubFile.aspx?id=13434&sid=indv.
Van Belle, E, Di Stasio, V, Caers, R, De Couck, M and Baert, S. 2018. Why are employers put off by long spells of unemployment? European Sociological Review, 34(6): 694–710. DOI: https://doi.org/10.1093/esr/jcy039
Van Oorschot, W. 2006. Making the difference in social Europe: Deservingness perceptions among citizens of European welfare states. Journal of European Social Policy, 16(1): 23–42. DOI: https://doi.org/10.1177/0958928706059829
Villadsen, AR and Wulff, AR. 2018. Is the public sector a fairer employer? Ethnic employment discrimination in the public and private sectors. Academy of Management Discoveries, 4(4): 429–448. DOI: https://doi.org/10.5465/amd.2016.0029
Wulff, JN and Villadsen, AR. 2019. Are survey experiments as valid as field experiments in management research? An empirical comparison using the case of ethnic employment discrimination. European Management Review. DOI: https://doi.org/10.1111/emre.12342
Wyszynski, MC. 2018. Good Refugees, Bad Migrants? Stereotype Content, Threat Perception and Helping Orientations towards Refugees, Migrants and Economic Migrants in Germany. University of Lisbon, Dissertation, Department of Social and Organizational Psychology. Available at https://repositorio.iscte-iul.pt/bitstream/10071/17870/1/Master_Mia_Caroline_Wyszynski.pdf.
Zschirnt, E and Ruedin, D. 2016. Ethnic discrimination in hiring decisions: A meta-analysis of correspondence tests 1990–2015. Journal of Ethnic and Migration Studies, 42(7): 1115–1134. DOI: https://doi.org/10.1080/1369183X.2015.1133279